Author: Dezhi Zhao
Date: 13:30:30 06/24/04
Go up one level in this thread
On June 24, 2004 at 16:00:16, Dieter Buerssner wrote: >On June 24, 2004 at 11:56:54, Dezhi Zhao wrote: > >>On June 24, 2004 at 11:17:50, Robert Hyatt wrote: >> >>>On June 24, 2004 at 09:03:57, Gian-Carlo Pascutto wrote: > >>>>2) You need to handle the branch too. The cost of that is going to be way >>>>higher than the cost of setting up/restoring a register. >>> >>> >>>That branch is probably predicted perfectly. it is "yes" "no" "yes" "no". >>>something the PIV handles well although others do not. > >I don't think, that it really matters (speed wise), but I would assume, that it >is not a simple yes no yes no sequence. You can hit qsearch, and no capture >moves are possible. You retract one move, and hit qsearch again, with the same >color to move. Both time eval is called with the same wtm. In other cases, your >yes no sequence will happen. But to me, it doesn't look very predictable at all. At least PIV can do the prediction though it may not be a simple pattern. I mean in the previous post that a CPU can lose its track of the mentioned branch if it meetsmuch code in bewteen. >Neither would I think, that it influences speed much (branch predicted correctly >or not). I think, like Gerd, the main issue is easiness of changing the code and >the chance to code some bugs. With basically duplicated code, the probability >for errors will be higher. partially agreed. However, you can still wrap it up by using inline functions, templates or even macros. > >I did some speed test for my incheck function. Coding it seperately for white >and black produces somewhat better code. In practice, I found no speed >difference. > >Regards, >Dieter
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.